Channel Estimation And Pilot Design For Massive MIMO Systems Based On Compressed Sensing

With the development of multimedia services,Internet of Things and big data,data transmission rate of the future mobile communication systems should be increased by tens of times or even thousands of times,and the existing communication technology can not meet this demand.Massive multiple-input and multiple-output(MIMO)is considered as a core technology of the next generation mobile communication systems,since it can make full use of the freedom of space and improves the data transmission rate of the systems.However,the performance of massive MIMO systems depends largely on the quality of the channel state information(CSI)obtained by the base station.Both pilot and feedback overhead required by conventional pilot-aided channel estimation methods increase with the number of antennas at the base station,and this will undoubtedly limit the performance gain brought by massive MIMO.Therefore,this thesis focuses on the efficient channel estimation scheme and pilot design.Firstly,this thesis investigates the traditional downlink channel estimation methods in MIMO systems,including least square(LS),scaled least square(SLS)and minimized mean square error(MMSE)method.The derivations of these channel estimation methods are described at great length firstly,then we analyze the optimal pilot structure and mean square error performance of each method.Theoretical results show the influence of the number of antennas at the base station on the error performance and the pilot overhead of these methods.Besides,we also introduce some improved and simplified methods for LS and MMSE in detail.Secondly,we propose the compressed sensing based channel estimation scheme for massive MIMO sys-tems in the beam domain in regard of huge pilot consumption and high computational complexity of conven-tional methods.Massive MIMO channel presents sparse characteristics in the beam domain,and the channel energy is mainly concentrated on a few beams.Therefore,this feature can be used to reduce the pilot overhead and system complexity.According to the sparse characteristics,this thesis proposes a beam domain channel estimation scheme based on compressed sensing,and gives a non-orthogonal pilot design which satisfies the restricted isometry property(RIP).Meanwhile,we propose a modified subspace pursuit(M-SP)algorithm to solve approximately sparse signal recovery problems.Simulation results show that the proposed method has better performance than traditional LS channel estimation,and can effectively reduce the pilot overhead.Finally,this thesis focuses on channel estimation and pilot design for massive MIMO-OFDM systems,and we propose the beam-delay domain channel estimation scheme based on compressed sensing for massive MIMO-OFDM systems.Massive MIMO-OFDM channels are sparse in many typical propagation scenarios,the channel energy is mainly concentrated on a few beams,and such sparsity can also be found in the delay domain which means that a small fraction of the channel tap coefficients occupies most of the channel energy.Therefore,we propose a beam-delay domain channel estimation scheme based on these sparse characteristics,and design the phase shift Zadoff-Chu pilots(PSZCPs)in the beam-frequency domain and derive the corre-sponding beam-delay domain measurement matrix.Besides,MMSE based SP(MMSE-SP)algorithm using statistical CSI is proposed to improve the performance.Simulation results show that the proposed beam-delay domain channel estimation can effectively reduce the pilot overhead and achieve significant performance gain over conventional channel estimation based on phase shift orthogonal pilots(PSOPs).Our method requires only one OFDM symbol,while the conventional methods require multiple OFDM symbols.In addition,the proposed MMSE-SP algorithm improves the performance of SP algorithm to a great extent,especially in the low signal-to-noise ratio(SNR)regime,and reduces the dependency on the precise sparsity at the same time.